Rural-to-urban migration and risk of hypertension: Longitudinal results of the PERU MIGRANT study

Abstract

Urbanization is a hazardous process because it affects health population due to changes in diet and physical activity patterns. This study aimed to determine the effect of migration on the incidence of hypertension. Participants of the PERU MIGRANT study, i.e. rural, urban, and rural-to-urban migrants were re-evaluated after five years from baseline. The outcome was incidence of hypertension; and the exposures were study group and other well-known risk factors. Incidence rates, relative risks (RR), and population attributable fractions were calculated. At baseline, 201 (20.4%), 589 (59.5%), and 199 (20.1%) were rural, rural-to-urban migrant and urban subjects, respectively. Overall mean age was 47.9 (SD±12.0) years, and 522 (52.9%) were females. Hypertension prevalence at baseline was 16.0% (95% CI 13.7%–18.3%), being more common in urban group; whereas pre-hypertension was more prevalent in rural participants (p<0.001). Follow-up rate at 5 years was 94%, 895 participants were re-assessed and 33 (3.3%) deaths were recorded. Overall incidence of hypertension was 1.73 (95%CI 1.36–2.20) per 100 person-years. In multivariable model and compared to the urban group, rural group had a greater risk of developing hypertension (RR=3.58; 95%CI 1.42–9.06). Population attributable fractions showed high waist circumference as the leading risk factor for the hypertension development in rural (19.1%), migrant (27.9%), and urban (45.8%) participants. Subjects from rural settings are at higher risk to develop hypertension relative to rural-urban migrant or urban groups. Central obesity was the leading risk factor for hypertension incidence in the three population groups.

Hypertension is a major problem worldwide, including low- and middle-income countries (LMIC) ¹. More than 60% of deaths due to non communicable diseases (NCDs) are attributable to preventable cardiometabolic factors, with high blood pressure enclosing the largest effect ^{2, 3}. Population aging, smoking, sedentary lifestyles, and dietary patterns, in the context of globalization and unplanned urbanization, are responsible for the high prevalence of hypertension ⁴.

The urbanization process is occurring at a fast rate, especially in LMIC ⁵. Rural-to-urban migration is one of the key drivers of the urbanization process in LMIC. Urbanization is considered a hazardous process because it affects health population due to changes in diet and physical activity patterns, with a consequent increase in obesity, type 2 diabetes, and cardiovascular disease ⁶. As a result, there is a need of better understanding the impact of urbanization on NCDs.

A previous systematic review found that most of the reports assessing the impact of migration on cardiovascular health were cross-sectional in nature ⁷, with scarce information being derived from longitudinal studies ^8–12. Moreover, these latter papers focused on changes in blood pressure instead of occurrence of hypertension over time. Given the prominence of migration and urbanization in LMIC as closely linked processes to NCDs, an additional limitation from available prospective designs relates to the low number of migrants evaluated. Peru offers a unique opportunity to assess the potential impact of within-country rural-to-urban migration on cardiovascular health: migration patterns changed during the political violence between 1970–1990's period ¹³, with several deaths and large amounts of displaced people ¹⁴. Therefore, migration was largely driven by the need of escaping from the armed conflict rather than economic reasons.

Given the aforementioned framework, we hypothesized that the rural-to-urban migrant group will have higher mean blood pressure, and for instance, higher rates of hypertension, than their rural peers, yet not as high as their urban counterparts. As a result, the aim of this study was twofold: first, to determine the effect of rural-to-urban migration on the incidence of hypertension; and second, to compare the role of potential risk factors on hypertension occurrence according to study population groups.

METHODOLOGY

Information comes from the first follow-up assessment of the PERU MIGRANT study ¹⁵, an ongoing prospective cohort designed to assess the magnitude of differences between rural, rural-to-urban migrant and urban groups in specific cardiovascular risk factors.

Two different settings were considered for this study. San Jose de Secce, a village located in Ayacucho was selected as the rural study site. Ayacucho, an Andean region, was one of the most affected areas during the period of conflict (1988–1993) in Peru ¹⁶. The area “Las Pampas de San Juan de Miraflores” in Lima, the capital of Peru, was selected as the urban area for the study. Both urban and rural-to-urban migrant participants were selected from this periurban area in the south of Lima.

At baseline, study groups were defined using a single random sampling in participants aged 30 years and over from the rural site of Ayacucho, the urban site of Lima, and rural-to-urban migrants from Ayacucho now residing in Lima. Information regarding selection criteria, sample size and participation rates have been published elsewhere ¹⁷. Rural and urban subjects had always lived in their respective settings. For this evaluation, participants were re-contacted in the same setting where they were originally enrolled. We did not collect information about moving back to rural areas. Since participants were re-contacted where they were at baseline, we assumed they did not move and particularly the migrant group did not go back to their rural birthplace.

The exposure of interest was study group, defined as rural, migrant and urban according to the baseline assessment. Other exposures assessed at baseline and part of this analysis were: binge drinking, defined as two or more nights of alcohol consumption in the month before the assessment and having ever drunk 6 or more drinks at a time; current daily smoking, defined as the self-report of smoking ≥1 cigarette per day; low physical activity levels defined in accordance with the International Physical Activity Questionnaire (IPAQ) protocol, thus, the categorical physical levels were coded based on total days of physical activity and metabolic equivalent (MET) in minutes/week ¹⁸; high total cholesterol, defined as fasting total cholesterol ≥200 mg/dL ¹⁹; obesity, defined as body mass index (BMI) ≥30kg/m²; high waist circumference, defined according to International Diabetes Federation cutoffs for South American populations ²⁰; and type 2 diabetes, defined as any of the following conditions: fasting glucose ≥126 mg/dL, self-report of physician diagnosis and currently receiving anti-hyperglycemic drugs ²¹. Baseline fasting blood samples were obtained and analyzed in a single facility and the quality of assays was checked with regular external standards and internal duplicate assays monitored by BioRad (www.biorad.com). Total cholesterol was measured on serum, whereas glucose was measured in plasma using an enzymatic colorimetric method (GOD-PAP, Modular P-E/Roche-Cobas, Germany).

Other variables of interest, also assessed at baseline, included as potential confounders were: sex, age (30–49 and 50+ years), education level (none/some primary school, complete primary, and some secondary), and socioeconomic status, using a wealth index based on assets and household facilities and categorized separately into tertiles for study group, and merged into a single variable ²².

The outcome of interest was the occurrence of hypertension, defined as the presence of high blood pressure levels (systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg) according to international guidelines ²³ or current use of anti-hypertensive medication prescribed by a physician. Blood pressure was assessed in triplicate, after a 5-minute resting period, using an automatic monitor OMROM HEM-780 (OMRON, Tokyo, Japan) previously validated for adult population ²⁴. In addition, pre-hypertension was accordingly defined as a systolic blood pressure from 120 to 139 mmHg or diastolic blood pressure from 80 to 89 mmHg. Participants originally enrolled in the PERU MIGRANT study, from 2007 to 2008, were re-contacted from 2012 to 2013 in the same setting where they were enrolled at baseline. After oral consent, they were requested to respond to a detailed questionnaire. Fieldworkers in rural areas were fluent in Spanish and Quechua, and they administered the survey to those with poor literacy. Weight and waist circumference were measured in triplicate by fieldworkers using standardized techniques.

Statistical analysis was conducted in STATA 13 for Windows (STATA Corp, College Station, TX, USA). Population characteristics were tabulated according to study group at baseline. Chi squared test was used to compared categorical variables, whereas continuous variables were compared using analysis of variance test. Incidence rates per 100 person-years of follow-up and 95% confidence intervals (95%CI) were calculated excluding those having hypertension at baseline. Incidence estimates were obtained by potential risk factors and study groups. Generalized linear models, assuming Poisson distribution, were utilized to determine the strength of association, i.e. relative risks (RR), between study groups and hypertension, controlling for several potential confounders. In addition, post-hoc analyses were also performed in the migrant group using only migration surrogates (age at migration and years lived in urban area). Crude and adjusted models were also generated to determine RR of well-established risk factors for hypertension by study group. Given the number of confounder variables, variance inflation factor was used to determine collinearity. Finally, population attributable fractions (PAF) were determined by using the punaf command for STATA ²⁵.

Ethical approval for baseline and follow-up phase was granted by the Institutional Review Board at Universidad Peruana Cayetano Heredia in Lima, Peru. Participants provided verbal informed consent due to major illiteracy rates, especially in rural areas.

RESULTS

At baseline, data from 988 participants was analyzed, mean age was 47.9 (SD: 12.0) years, 522 (52.8%) females, and according to study group, 201 (20.4%), 589 (59.5%), and 199 (20.1%) were rural, rural-to-urban migrant and urban participants, respectively. Regarding the migrant group, mean age at first migration was 14.7 (SD: 9.0) years; in addition, mean time lived in an urban area was 32.0 (SD: 10.5) years. Characteristics of the study population according to study group are shown in Table 1. Overall prevalence of hypertension was 16.1% (159/988; 95%CI 13.8%–18.4%) and prevalence estimates varied by study group: hypertension was more common among urban population, but pre-hypertension was more prevalent among rural group (p<0.001). Moreover, of all participants with hypertension, 50% of rural dwellers, 77.6% of migrants, and 47.5% of urban individuals were previously diagnosed by a physician (p = 0.001).

Table 1.

Baseline characteristics of the study sample according to study group

		Rural group (n = 201)	Migrant group (n = 589)	Urban group (n = 199)	p-value
Sex	Male	95 (47.3%)	280 (47.5%)	92 (46.2%)	0.95
Age	30–49 years	117 (58.2%)	337 (57.2%)	110 (55.3%)	0.57
	50+ years	84 (41.8%)	252 (42.8%)	89 (44.7%)
Education level	None/some primary school	132 (65.7%)	183 (31.1%)	13 (6.6%)	< 0.001
	Complete primary	30 (14.9%)	99 (16.8%)	23 (11.6%)
	Some secondary	39 (19.4%)	306 (52.1%)	162 (81.8%)
Socioeconomic status	Lowest tertile	196 (97.5%)	119 (20.2%)	32 (16.1%)	< 0.001
	Middle tertile	5 (2.5%)	253 (43.0%)	69 (34.7%)
	Highest tertile	0 (0.0%)	217 (36.8%)	98 (49.2%)
Binge drinking	Yes	23 (11.4%)	47 (8.0%)	17 (8.5%)	0.32
Current daily smoking	Yes	1 (0.5%)	15 (2.6%)	17 (8.5%)	< 0.001
Physical activity	Low levels	4 (2.05)	173 (29.7%)	78 (39.4%)	< 0.001
Total cholesterol	≥ 200 mg/dL	15 (7.5%)	220 (37.4%)	71 (35.7%)	< 0.001
High waist circumference	Yes	30 (15.2%)	354 (60.3%)	132 (66.7%)	< 0.001
Obesity	BMI ≥ 30 kg/m2	6 (3.0%)	124 (21.1%)	68 (34.2%)	< 0.001
Type 2 diabetes	Yes	3 (1.5%)	21 (3.6%)	16 (8.0%)	0.003
Systolic blood pressure (mmHg) *	Mean (SD)	120.9 (18.7)	119.9 (16.4)	128.2 (22.9)	< 0.001
Diastolic blood pressure (mmHg) *	Mean (SD)	74.2 (9.2)	71.3 (9.3)	76.2 (11.5)	< 0.001
Blood pressure status	Normal	106 (52.7%)	318 (54.1%)	87 (43.7%)	< 0.001
	Pre-hypertension	71 (35.3%)	194 (33.0%)	53 (26.6%)
	Hypertension	24 (11.9%)	76 (12.9%)	59 (29.7%)

Results may not add due to missing values.

^*Analysis of variance test was used for comparisons, instead of Chi squared test for categorical variables

Of the 988 participants enrolled at baseline, 60 (6.1%) were lost to follow-up, and 33 (3.3%) died. Thus, of the 895 (90.6%) re-contacted, 133 (14.9%) were further excluded from incidence calculations because hypertension diagnosis at baseline. Mean time of follow-up was 5.2 (SD: 0.6) years, completing a total of 3,962 person-years of follow-up. A total of 66 new cases of hypertension were found with an overall incidence of 1.73 (95%CI 1.36–2.20) per 100 person-years (5-year cumulative incidence: 8.65%). At follow-up, systolic and diastolic blood pressure means were similar according to study group (Online supplement, E-Table 1). According to study group, hypertension incidence was 2.44 (95%CI 1.62–3.67) in the rural group, 1.60 (95%CI 1.15–2.22) in the rural-to-urban migrant group, and 1.11 (95%CI 0.53–2.33) in the urban group (p<0.001). In addition, of all new cases, 91.3% of rural dwellers, 75.0% of migrants, and 100% of urban individuals reported to be diagnosed of hypertension by a physician in the previous five years of the follow-up (p = 0.19). Hypertension incidence according to population characteristics and study group at baseline are shown in Table 2. Of note, incidence of hypertension due to pre-hypertension was highest among rural than migrant or urban participants.

Table 2.

Incidence rates and 95%CI of hypertension according to population characteristics at baseline

		Rural group	Migrant group	Urban group
Sex	Female	2.65 (1.57 – 4.48)	2.14 (1.46 – 3.14)	1.72 (0.77 – 3.83)
	Male	2.17 (1.13 – 4.18)	0.97 (0.52 – 1.80)	0.35 (0.05 – 2.52)
Age	30–49 years	1.98 (1.12 – 3.49)	1.07 (0.65 – 1.77)	0.95 (0.36 – 2.54)
	50+ years	3.27 (1.81 – 5.91)	2.49 (1.62 – 3.82)	1.42 (0.46 – 4.39)
Education level	None/some primary school	2.23 (1.30 – 3.85)	2.48 (1.52 – 4.05)	--
	Complete primary	2.47 (0.93 – 6.58)	1.36 (0.56 – 3.26)	2.22 (0.31 – 15.78)
	Some secondary	3.03 (1.36 – 6.75)	1.22 (0.74 – 2.03)	1.10 (0.49 – 2.45)
Socioeconomic status	Lowest tertile	1.95 (1.08 – 3.52)	1.57 (0.93 – 2.65)	1.82 (0.68 – 4.84)
	Middle tertile	1.52 (0.21 – 10.76)	1.96 (1.11 – 3.45)	0.53 (0.075 – 3.76)
	Highest tertile	3.53 (1.95 – 6.37)	1.35 (0.73 – 2.50)	0.90 (0.23 – 3.60)
Binge drinking	No	2.42 (1.56 – 3.74)	1.68 (1.21 – 2.34)	1.22 (0.58 – 2.55)
	Yes	2.63 (0.85 – 8.16)	0.60 (0.08 – 4.25)	--
Daily smoking	No	2.35 (1.55 – 3.57)	1.65 (1.19 – 2.28)	1.18 (0.56 – 2.48)
	Yes	16.67 (2.35 – 118.32)	--	--
Physical activity	Moderate/high levels	2.47 (1.64 – 3.72)	1.67 (1.14 – 2.45)	1.32 (0.55 – 3.18)
	Low levels	--	1.52 (0.82 – 2.82)	0.79 (0.20 – 3.16)
Total cholesterol	< 200 mg/dL	2.18 (1.39 – 3.42)	1.31 (0.84 – 2.06)	1.02 (0.38 – 2.73)
	≥ 200 mg/dL	5.56 (2.09 – 14.80)	2.13 (1.32 – 3.42)	1.25 (0.40 – 3.88)
High waist circumference	No	2.05 (1.26 – 3.35)	0.89 (0.45 – 1.79)	0.41 (0.06 – 2.93)
	Yes	4.67 (2.22 – 9.79)	2.08 (1.44 – 3.01)	1.56 (0.70 – 3.48)
Obesity	BMI < 30kg/m2	2.32 (1.51 – 3.55)	1.45 (0.99 – 2.13)	0.45 (0.11 – 1.82)
	BMI ≥ 30 kg/m2	5.56 (1.39 – 22.21)	2.19 (1.18 – 4.07)	2.62 (1.09 – 6.29)
Type 2 diabetes	No	2.47 (1.64 – 3.72)	1.52 (1.08 – 2.14)	0.85 (0.35 – 2.05)
	Yes	--	3.95 (1.27 – 12.24)	4.55 (1.14 – 18.17)
Pre-hypertension	No	1.74 (0.93 – 3.23)	0.99 (0.58 – 1.67)	0.76 (0.24 – 2.34)
	Yes	3.55 (2.06 – 6.12)	2.66 (1.75 – 4.04)	1.71 (0.64 – 4.55)

Incidence rate was not calculated as there were no hypertension cases during follow-up (--)

Having the urban group as reference, rural participants had higher risk of hypertension and the magnitude of RR increased with further adjustment. After controlling for demographic and behavioral confounders, and compared to the urban group, rural participants were four times more likely to have hypertension (RR = 3.58; 95%CI 1.42 – 9.06). The migrant group was also at high risk of hypertension; however, results were not significant. Details are shown in Table 3. Using data from migration surrogates and after controlling for several confounders, those migrants living 30 years or more in the urban setting were at lower risk of hypertension when compared to those living less than 30 years; however, results were not significant (RR = 0.85; 95%CI 0.35 – 2.03). Similarly, when the age at migration was used, those who reported migrating at 15 years or over were at greater risk of developing hypertension when compared to those migrating at age below 15 years but results were not significant (RR = 1.04; 95%CI 0.53 – 2.04).

Table 3.

Relative risks of the association between hypertension and study group: crude and adjusted models

Study group	Crude model RR (95%CI)	Adjusted model* RR (95%CI)	Adjusted model** RR (95%CI)
Urban	1 (Reference)	1 (Reference)	1 (Reference)
Migrant	1.44 (0.66 – 3.17)	1.43 (0.65 – 3.15)	1.56 (0.73 – 3.30)
Rural	2.20 (0.97 – 4.97)	2.30 (0.93 – 5.71)	3.58 (1.42 – 9.06)

Bold estimates are statistically significant (p<0.05)

^*Adjusted by sex, age, education level, and socioeconomic status

^**Adjusted by sex, age, education level, socioeconomic status, binge drinking, current daily smoking, physical activity, high total cholesterol, obesity, high waist circumference, and type-2 diabetes.

As shown in Table 4, in multivariable models, of all risk factors explored, current daily smoking (RR 4.26) and high waits circumference (RR 2.68) were found to be associated with increased risk of hypertension among the rural group; and only pre-hypertension increased the risk among migrant population (RR 2.98). In the same vein, only type 2 diabetes at baseline (RR 7.10) increased the risk of hypertension among urban population.

Table 4.

Factors and risk of hypertension according to study group: crude and adjusted models

	Rural group		Migrant group		Urban group
	Crude model RR (95%CI)	Adjusted model* RR (95%CI)	Crude model RR (95%CI)	Adjusted model* RR (95%CI)	Crude model RR (95%CI)	Adjusted model* RR (95%CI)
Binge drinking
Yes	1.09 (0.36 – 3.33)	1.33 (0.43 – 4.14)	0.36 (0.05 – 2.52)	0.57 (0.08 – 4.11)	--	--
Current daily smoking
Yes	7.09 (4.81 – 10.5)	4.26 (1.44 – 12.5)	--	--	--	--
Physical activity
Low levels	--	--	0.91 (0.45 – 1.84)	0.87 (0.44 – 1.71)	0.60 (0.12 – 3.00)	0.90 (0.20 – 4.05)
High total cholesterol
≥ 200 mg/dL	2.54 (1.02 – 6.30)	2.34 (0.84 – 6.52)	1.63 (0.87 – 3.04)	1.27 (0.65 – 2.47)	1.22 (0.28 – 5.27)	1.05 (0.24 – 4.55)
Obesity
BMI ≥ 30 kg/m2	2.40 (0.72 – 7.98)	2.39 (0.59 – 9.68)	1.51 (0.76 – 3.02)	1.15 (0.57 – 2.34)	5.76 (1.16 – 28.7)	3.79 (0.83 – 17.3)
High waist circumference
Yes	2.28 (1.04 – 4.97)	2.68 (1.22 – 5.89)	2.33 (1.09 – 5.00)	1.56 (0.69 – 3.55)	3.78 (0.46 – 30.9)	2.15 (0.42 – 10.9)
Type 2 diabetes
Yes	--	--	2.60 (0.91 – 7.46)	1.70 (0.59 – 4.86)	5.34 (1.16 – 24.5)	7.10 (1.56 – 32.3)
Pre-hypertension
Yes	2.05 (0.96 – 4.38)	2.24 (1.00 – 5.02)	2.69 (1.41 – 5.12)	2.98 (1.55 – 5.73)	2.26 (0.52 – 9.76)	2.98 (0.74 – 12.1)

Bolded estimates are significant, p<0.05

^*Adjusted by sex, age, education level, and socioeconomic status.

Population attributable fractions were also calculated by study groups (Online Supplement E-Table 2) and showed considerable heterogeneity in terms of its magnitude between study groups. Many of the estimates were under 10% among rural and migrant groups, except high waist circumference (19.1% and 27.9%, respectively) and pre-hypertension (31.3% and 40.6%, respectively). On the other hand, obesity, high waist circumference, type 2 diabetes, and pre-hypertension were markedly important in the urban group (52.6%, 45.8%, 24.5%, and 37.9%, respectively).

DISCUSSION

This prospective ongoing cohort study included different study groups and was explicitly designed to ascertain whether differential risks for NCDs exist around rural-urban migrant and non-migrant groups. The risk of hypertension was almost four times greater among rural subjects relative to their urban counterparts. Although the migrant group had also an increased risk compared to urban individuals, this was not significant. Factors associated with hypertension incidence in the multivariate model differed by study group: high waist circumference and daily smoking in rural, pre-hypertension in rural-to-urban migrant, and type 2 diabetes in urban group. In addition, using estimates of population attributable fractions, obesity-related markers (i.e. body mass index, but especially high waist circumference) were the leading factors increasing the risk of hypertension in the three population group but particularly in urban individuals.

Our results are consistent with a higher risk of hypertension among rural when compared to urban group. This observation is supported by the high proportion of rural individuals found with pre-hypertension at baseline, but also for the high incidence rate of hypertension during follow-up. Although hypertension was markedly higher in the urban group, at baseline, a third of the rural and migrant populations had pre-hypertension. This information highlights the diversity of scenarios where hypertension evolves, signaling major risks in specific groups. This approach of identifying different risk magnitudes in low-resource settings suggests that other non-communicable conditions may also portray similar complexities, particularly in many LMIC in (epidemiological and nutritional) transition.

High pre-hypertension rates at baseline in our rural group as well as low health standards associated with living in rural settings, i.e. poverty, malnutrition, poor hygiene as well as inadequate health care ²⁶, might be potential explanations for these findings. Thus, these broader contextual variables applicable to rural settings, paired with ongoing nutritional transition, may potentially increase the risk of fast acquiring negative lifestyles profiles, especially obesity and then hypertension ²⁷. In addition, despite of removing the effect of different sociodemographic and lifestyles factors, our study lacked information about dietary patterns, a key determinant of cardiovascular outcomes ²⁸. Diet in population from Andean regions, as involved in this study, has changed over time. Energy from carbohydrates has had moderate declines regarding cereals especially, but sugars, as energy contributors have increased ²⁹. In addition, cholesterol intake has also increased, whereas vegetables, starchy roots and fruit intake have considerably reduced. As people living in rural areas had much higher levels of physical activity as shown in this study and a previous report ³⁰, we believe much of the effect seen in this study might be explained by diet. However, further studies are needed to evaluate the impact of diet on hypertension in these study groups.

Since the urbanization phenomenon in rural areas dates back to the last decades, another possible explanation for the higher hypertension incidence in the rural group is a period effect: higher incidence of hypertension in this population is a rather new feature. This is further supported as rural subjects had higher prevalence of pre-hypertension. Consequently, current urbanization process is reaching out to those pre-hypertensive subjects. This gives a window opportunity to implement prevention strategies before all pre-hypertensive rural subjects meet criteria for hypertension.

Early reports on this topic were performed in 1990s in Kenya and China ^{9, 31}, showing an increment of blood pressure due to rural-to-urban migration. These findings are contrary to our results perhaps because these past studies were conducted when travel and communication between rural and urban areas were more difficult and urban lifestyle were less likely adopted by rural dwellers. As a result, there is still limited longitudinal information available regarding the impact of migration on hypertension as many of the published studies report changes in blood pressure means instead of hypertension rates ^{8, 10, 11} and compared migrants versus non-migrant groups, instead of urban, rural-to-urban migrants, and rural populations. Thus, our findings expand on previous knowledge demonstrating a higher hypertension risk among rural population. Migrant populations have been thought to be potentially more affected by unhealthy practices acquired from living in a new setting, especially among those more acculturated ³². However, findings were always controversial. For example, a previous study suggested that some migration surrogates may directly influence broader social determinants of disease ³³, and hence, reduce the possibility to acquire cardiovascular outcomes. Many of these findings, nevertheless, come from cross-sectional studies and involved migrants who moved from developing to developed countries instead of rural-to-urban within-country migrants. In a post-hoc analysis, we tried to model the effect of migration surrogates (age at migration and years of urban exposure) but results were not conclusive.

In terms of blood pressure as continuous variable, previous cross-sectional studies have shown non-significant differences in blood pressure levels among migrants when compared to rural groups ^{15, 31}, and a pattern of significantly lower systolic and diastolic blood pressure in migrant compared to urban groups. Our study found differences in mean systolic and diastolic blood pressure levels by study group at baseline, but not after 5 years of follow-up. Two different longitudinal studies have reported changes in blood pressure levels in migrant groups ^{9, 10}. One of them found systolic blood pressure of migrants significantly higher that rural controls ⁹; whereas the other one found that both systolic and diastolic blood pressure were higher compared to non-migrants, but only in men ¹⁰.

Although cardiovascular risk factors are well known, global policies require local adaptation according to population profiles. For priority purposes, it is important to understand the local burden of disease including within-country heterogeneity of NCD distribution and their risk factors. For example, considering benefits and feasibility, reduction in tobacco is recommended as one of the best initiative ³⁴, but in our study, a small proportion (<5%) were daily smokers and, thus, the potential impact of an intervention focused on this risk factor would be almost negligible as highlighted by results of population attributable fractions.

According to our results, special attention needs to be paid to obesity, as BMI and waist circumference, showed a different distribution according to study group. In addition, type 2 diabetes was also another key factor among urban participants and a natural consequence of the increasing burden of obesity. Our results are compatible with the need of a reduction of central obesity ¹⁰, which might reduce, more than any other factor, the risk of hypertension in the three populations, but especially among urban people. These results can also reflect the fact that different regions and populations within a same country are at different stages of the nutritional transition. Therefore, interventions to prevent NCD increase in our context should be focused on reduction of obesity rates by improving the balance of intake (diet) and expenditure (physical activity) of energy.

The case of pre-hypertension deserves more attention as was present in almost one third of all the study groups at baseline. This is probably the stage at which interventions need to be more incisive and extensive to tackle the progress towards hypertension ³⁵. According to population attributable fraction, hypertension would be reduced in 30% in all study groups if a reduction of blood pressure under 120/80 mmHg was guaranteed.

Strength of this study includes the calculation of hypertension incidences over a 5 years period in well-defined rural, migrant and urban populations. Although the fact that our study population migrated due to different reasons than socioeconomic mobility is a novelty of the study, it could also be regarded as a limitation in terms of external generalizability. Migration is mostly driven by socioeconomic upward mobility, which comes along with different risk factors (e.g. sedentarism or unhealthy diets). Our study population did not move because they had achieved better socioeconomic standards, and thus our results may not fully represent the new migration waves across the world. Other limitations should also be described. First, although the rate of attrition during follow-up was relatively low (<10%), results might be affected by selection bias, especially in the urban setting for data collection involving both, migrants and urban subjects. As previously published ¹⁷, response rates of the baseline study were low in migrants (77.7%) and urban (56.8%) when compared to rural group (84.8%), and therefore participants originally enrolled might have other characteristics compared to those who rejected to participate. Second, although the definition of rural, migrant or urban population can change over time, we assumed this was not the case and did not affect our estimations as all of the participants were re-contacted in the same area where originally were enrolled. Third, power could be an issue as many well-recognized factors were not associated with the progression towards hypertension. However, as population attributable fractions assess the contribution of a risk factor to a disease, they can provide a better understanding of the role of these factors in the study populations. Finally, results could also reflect the effect of unmeasured confounders like chronic kidney disease. Unfortunately, we did not collect data about this condition at baseline; yet at follow-up we asked if the participant has been diagnosed with chronic kidney disease in the past five years showing a prevalence of less than 1%. In addition, a previous report in the urban study area found that about 20% of the population presented some degree of chronic kidney disease, and 19% of subjects with chronic kidney disease had hypertension as well ³⁶. Despite of these findings our results showed such a strong hypertension risk given the different population groups included as the exposure variable, that it is hard to think the risk would be completely explained by chronic kidney disease. The extent at which chronic kidney disease confound, or explain, the association of interest must be addressed by future studies, particularly as the burden of chronic kidney disease is rather neglected in rural resource-limited settings.

In conclusion, the incidence of hypertension was high in rural populations when compared to migrant and urban groups. Risk factors for hypertension were different according to study group, and almost one third of the three populations had pre-hypertension at baseline. Obesity, assessed by waist circumference and body mass index, was the leading risk factor for hypertension in the three groups evaluated. Results suggest that interventions to tackle hypertension need to focus in the reduction of obesity, especially in urban settings.

SUMMARY TABLE.

What is known about the topic

• Urbanization affects health population due to changes in diet and physical activity patterns, thus increasing cardiovascular diseases (CVD).

• Impact of migration on CVD and risk factors, such as hypertension, is derived from cross-sectional studies only.

What this study adds

• In this prospective study, the risk of developing hypertension was about four times higher in rural compared with urban individuals.

• After 5.2 years of follow-up, migrating from rural to urban areas did not carry significant risk of hypertension compared to stay living in urban areas.

• Obesity, as waist circumference and body mass index, was the leading risk factor for hypertension development in the three study groups.